[DMP 2026]: Building Multilingual Voice Agents to Improve Learning Engagement in Government School Systems #4
Description
Ticket Contents
Description
Student drop-off after onboarding is one of the biggest challenges in large-scale government deployments of TAP's learning platform. Contributing factors include limited digital familiarity, lack of parental awareness, inconsistent school follow-up, and low motivation without reminders. This project involves building AI-powered multilingual voice agents that proactively call students and families, guiding them back into their learning journeys in local languages (Hindi, Marathi, Punjabi). Unlike SMS or WhatsApp reminders, voice conversations allow contextual nudging, task explanation, and personalized interaction, making them significantly more effective for re-engagement.
Goals & Mid-Point Milestone
Goals
- Design a scalable voice agent architecture supporting automated outbound calls via phone and WhatsApp
- Implement multilingual interaction with automatic language detection (Hindi, Marathi, Punjabi)
- Build a learning nudging engine integrated with TAP's LMS to personalize calls based on student activity
- Design structured conversational flows with fallback and escalation logic
- Implement an experimentation framework to measure impact on engagement metrics (return-to-learning rate, completion rate, parent engagement)
- [Goals Achieved By Mid-point Milestone]: Working voice agent prototype capable of initiating automated outbound calls in at least one language with a basic conversational nudging flow connected to the LMS
Setup/Installation
No response
Expected Outcome
A working multilingual voice agent prototype
Integration with TAP’s learning management system
Ability to initiate automated calls and handle simple conversations
Language detection and response in supported languages
Documentation for scaling the system in government deployments
Acceptance Criteria
No response
Implementation Details
- Voice agent framework: VAPI or similar voice orchestration platform
- Calling channels: standard telephony APIs + WhatsApp calling/voice integration
- Multilingual STT/TTS pipelines for Hindi, Marathi, and Punjabi
- LMS integration via TAP's Frappe backend REST APIs to retrieve student activity and progression data
- Conversational flow design: structured dialogues, fallback logic, escalation paths
- Experimentation infrastructure for measuring and logging engagement impact metrics
Mockups/Wireframes
No response
Product Name
Multilingual Voice Agent for Student Engagement and Learning Nudges
Organisation Name
The Apprentice Project
Domain
Education
Tech Skills Needed
Artificial Intelligence, Natural Language Processing, WebSockets, RESTful APIs
Mentor(s)
TBD
Category
AI